Retinex-Based Low Contrast Image Enhancement Using Adaptive Tone-Mapping

Abstract

In this paper, we enhance low contrast images using the human visual system based Retinex theory and adaptive tone-mapping. We try to reduce halo artifact and color inconsistency, but also preserve naturalness of images. In the proposed algorithm, we process only the Y channel of the Yuv color space rather than RGB color space to maintain color-constancy. We first apply an adaptive bilateral filtering on the Y channel image to alleviate halo artifact during enhancement. Then we partition the intensity range of probability distribution of filtered Y channel image into low, middle, and high contrast regions according to a cost function. We improve the contrast of filtered Y channel image by using A-law based tone mapping by stretching the low contrast regions and compressing the high contrast regions adaptively. Experimental results show that the proposed algorithm enhances the visibility of input low contrast images efficiently.